What’s with the historical trend juxtapositions?

When is it OK to juxtapose historical trends?

You have to watch out for this (via Boing Boing):

There is no reason to suspect that the rise in autism is linked to the rise in organic food sales.

But other times it seems reasonable to me, like this:


There are lots of reasons the long-run decline in fertility is related to the rise in women’s employment rates. We know from lots of research that women with more children are less likely to have jobs; women with jobs are less likely to have children; and over time the proportion of women in the second group has grown relative to the first.

So it’s OK to use eyeballed historical trends when you have good research backing up the association. Your conclusions, then, don’t rely on the simple trend comparison. The trends are an illustration.

But trends need not have have a simple cause-and-effect relationship — or a unidirectional relationship — for it to be important to compare them. Sometimes the relationship is just descriptively important. So, looking at the graph above, it would be reasonable to say, “Women’s lives sure have changed. They have fewer children and more jobs, on average, than they used to.”

And then there is the negative case. The other day I complained when Kay Hymowitz implied that the rise in father-absent families caused an increase in crime among boys. And I offered this simple trend comparison to undermine that story:

It was not my intention to say there is no connection between father absence and boys’ criminal behavior. (I’ve sketched out some possible links in this old post; and made essentially the same comparison about single mothers and crime before.) But the lack of a strong correlation in the trends over time is a challenge. That’s what I’ve been arguing about cell phones and traffic accidents:

Of course driving and texting is dangerous. And of course single parents have a harder time (on average) supervising and disciplining their children than married parents. But if there is a big discordance between the trends — texting and driving, single parents and crime — then that’s a problem for the story that one trend is driven by the other. Causal relationships may be apparent in a lab, or at the margins, but to explain large-scale social change is more difficult — and that’s often what we’re trying to do when we draw from specific research to make political, policy, or theoretical arguments.

So, it’s OK to use discordant trends to take potshots at a proposed causal story, to express skepticism. The discordant trends are a hurdle for the theory to overcome.

If you have good research showing that single parenthood, and especially father absence, is harming boys more than girls, then it would be to OK to use trends as an illustration. It just can’t be your main evidence. So Kay Hymowitz could reasonably include a graph like this to accompany her extensive review of the research on family structure and trouble for boys:


Yes, women’s advantage in high school and college completion has accompanied the trend toward father-absent living arrangements for young boys. That doesn’t fulfill her need to present more direct evidence, but it’s a piece of supporting evidence.

Conclusion: Juxtaposing historical trends is not how you prove a theory. It is a great tool for illustrating known associations, for describing social change, and for challenging theories or narratives.

7 thoughts on “What’s with the historical trend juxtapositions?

  1. Would the lead in blood versus crime not constitute a historical trend juxtaposition that you commented on earlier? There is no scientific study that shows lead in the blood of a single person leads to crime although it shows it leads to brain dysfunction.


  2. This is a neat overview of the issues around correlating two trends. I think I will assign it to my students.

    One rule of thumb in evaluating trends is to be especially skeptical of “evidence” based on two trends that are consistently linear. This catches the problem in your autism & organic foods correlation for instance. To be useful, the trends really should change direction or at least level off at some point so that you can test whether the two trends show similar changes.

    That test leads one to question the importance of fertility decline in the long-run rise (and recent plateau) in women’s labor force participation. True, we have a lot of micro-evidence that motherhood reduces employment for individual women (& vice-versa). But the macro-evidence is quite weak and the two trends graphed here illustrate those problems. First, of course, is the baby boom of the 1950s+ which seems to have no noticeable impact on the rise of women’s labor force participation. Second, the decline in the fertility rate stalls around 1970 while the rise in women’s labor force participation doesn’t stall until the mid 1990s. A clever argument might explain why there was a quarter century lag in those two changes, but our initial reaction should be skepticism that these two rates have much of a causal relationship.


  3. This summer I gave a lecture at the Summer Workshop of the Luxembourg Income Study (LIS) that started with the correlation between total fertility rates (TFR) and female labor force participation rates (FLFP), similar to the second figure in this post. Similar to Reeve, I am quite hesitant to interpret this correlation, but on a different (additional) ground.

    As the cross-country correlation between FLFP and TFR turned from negative to positive around 1985, this as often interpreted as the result of family policies allegedly having made motherhood and employment easier to combine. This may very well be true, but I maintain this cannot be directly inferred from the country-level correlation. Instead, research on the compatibility of motherhood and employment should use person-level data.

    Some slides from the lecture are available here: https://www.researchgate.net/publication/250310261_Family_Policies_Women%27s_Earnings_and_Between-Household_Inequalities_Using_LIS_for_comparative_analyses?ev=prf_pub


    1. Thanks for sharing your slides. Do you have a paper?

      I’m quite confident of the relationship over the long run — as in, olden days versus nowadays. But clearly in late-modern rich countries the situation is different.


      1. I agree with you that the increased female labor force participation rates to an important extent driven by decreased fertility. In that sense, the second figure in this post makes complete sense, particularly because we have an understanding of the micro-level mechanism: mothers are less likely to be employed than women without children (the reverse also holds).

        I was referring to the cross-country correlation between TFR and FLFP (indeed, similar but not the same as the figure here), which has turned from negative to positive in OECD countries between 1970 and 2000. The reversal of this correlation between TFR and FLFP has been observed by various authors, including:

        Ahn, N., & Mira, P. (2002). A note on the changing relationship between fertility and female employment rates in developed countries. Journal of Population Economics, 15(4), 667–682. doi:10.1007/s001480100078

        Kogel and colleagues used country-level data to analyze the changing correlation, and point out that it was not likely that on the person-level the correlation between fertility and employment had not changed its sign:

        Kögel, T. (2004). Did the association between fertility and female employment within OECD countries really change its sign? Journal of Population Economics, 17(1), 45:65.

        Not being able to use person-level data, however, Kögel et al. could not directly observe the person-level correlation. A notable attempt to do so was made by Matysiak & Vignoli, presenting a meta-analysis of person-level studies.

        Matysiak, A., & Vignoli, D. (2008). Fertility and Women’s Employment: A Meta-analysis. European Journal of Population, 24, 363–384.

        In my own study, I built upon this literature using harmonized person-level data to describe and analyze the person-level correlation between motherhood and employment in 18 OECD countries covering 1975-1999.

        Nieuwenhuis, R., Need, A., & Van der Kolk, H. (2012). Institutional and Demographic Explanations of Women’s Employment in 18 OECD Countries, 1975-1999. Journal of Marriage and Family, 74(June), 614–630. doi:10.1111/j.1741-3737.2012.00965.x

        (This latter paper can also be found here:



Comments welcome (may be moderated)

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s